This is the MATLAB source code for reproducing the results in the paper entitled "Haziness Degree Evaluator: A Knowledge-Driven Approach for Haze Density Estimation", submitted to the MDPI Sensors journal.
The source code of the proposed HDE is placed in the folder "source_code" under the filename "new_indicator_v5_opt.m". In addition, we also provided our implementation of the work of Jiang et al. under the filename "ref_model.m".
In the "data" folder, we provided the ".mat" files that contain the FADE, DF, and HDE values calculated on the two testing sets of the RESIDE dataset. Therefore, it would be more convenient to leverage those results instead of calculating them again.
To reproduce the results on hazy/haze-free image classification on new data, please run the script "acc_test.m". The results would be identical those in the following table.
Class | FADE | DF | HDE |
---|---|---|---|
DV | 0.9866 | 0.2968 | 0.8811 |
P | 1020 | ||
TP | 863 | 242 | 929 |
TPR | 84.6% | 23.7% | 91.1% |
FN | 157 | 778 | 91 |
FNR | 15.4% | 76.3% | 8.9% |
N | 552 | ||
TN | 446 | 508 | 507 |
TNR | 80.8% | 92.0% | 91.1% |
FP | 106 | 44 | 45 |
FPR | 19.2% | 8.0% | 8.1% |
ACC | 83.3% | 47.7% | 91.4% |
Furthermore, the run-time comparison results can also be reproduced by invoking the script "runtime_meaaure.m" on test images in the "test-images-for-runtime-measure" folder. For executing the script, it is required to specify the path to the FADE's source code, which can be obtained from http://live.ece.utexas.edu/research/fog/index.html. Also, please specify the path to test images on line 21.